Somebody is handed transport. Sometimes it is a new plant, sometimes a company that has always let the sales team book trucks and has decided to stop. The first thing that happens is almost always the same: they ask for quotes, or they price trucks.
Both are answers. The problem is that neither question can be answered yet, so whatever comes back gets treated as a fact, and the rest of the operation is built to fit it.
There is an order to these decisions, and it is not the order they arrive in. Each one is an input to the next, and they get cheaper to change as you go down the list. This is that order, with what each step needs before it can start and what it costs to undo later.
Why the order is the whole problem
Transport is the largest single block of the national logistics bill. NESDC put Thailand's logistics costs at 2,509.4 billion baht in 2024, or 13.5% of GDP, split into transportation at 47.9%, inventory holding at 44.7% and administration at 7.4%.
That first number is what people set out to manage. The trap is that most of it is decided by things that are not transport decisions at all: what you promised customers, where you ship from, and what shape your orders arrive in. Get those settled in the wrong sequence and the transport plan is just an expensive way of absorbing them.
The reversal costs run in one clear direction. A route can be changed on a Tuesday. A delivery promise sits in customer contracts. A site sits in a lease. So the decisions that are hardest to unwind have to be made first, on the best information you can get, and the ones that are easy to unwind should be left until you actually know something.
Decision 1: what you already ship
Nothing below this line can be decided without it, and it is the step that gets skipped.
Pull eight to twelve weeks of despatch history into one table. Not a forecast, not a typical week somebody remembers. Every consignment that actually left, with five columns:
- the date it went
- the delivery address
- the weight
- the volume, or the pallet or carton count if that is what you have
- the delivery window the customer asked for
Eight weeks is the minimum that survives one bad week. Twelve is better if month-end distorts you, which it usually does.
That table is not a report. It produces three distributions, and each one is the input to a decision further down:
- Drop size. How big is a typical delivery, and how wide is the spread. This decides the vehicle profile.
- Location repeat rate. How many delivery points appear in most weeks against how many appear once. This decides whether the work can be run on a fixed frame at all.
- Day shape. How many drops on the busiest day against the average day. This decides how much capacity you commit and how much you buy when you need it.
The cost of skipping it is not theoretical. In a Thai case published in 2025, a distribution operation's high transport cost was traced by cause-and-effect analysis not to prices but to the absence of a systematic planning process. Rebuilding the plan from the company's own existing customer and demand data took monthly trips from 87 to 49 and average monthly transport cost from 162,500 baht to 91,500 baht, a fall of 71,000 baht or 43.69%. The before figure averages four months and the after figure two, so treat the size with some care. The direction is the point: the money was in the planning, and the data needed to do the planning was already in the building.
Decision 2: what you promise
Order cut-off, lead time, delivery frequency, minimum order and the delivery windows you accept. These look like sales decisions and they are made by sales, usually years ago, usually in one afternoon.
They are also the transport budget. They fix how many stops you buy, how full each vehicle leaves and whether a planner has anything to work with. That machinery is set out in the three order rules that set your transport cost, and it is worth settling before anything else because it is the second most expensive thing on this list to reverse. A promise lives in customer contracts and in customer expectations, and taking one back is a commercial negotiation, not an operational change.
At setup you have an advantage you will never have again: nothing has been promised yet. Use it. Decide the cut-off, the standard lead time and the minimum drop before the first customer is told anything, and write down which of them you are willing to sell against later.
Decision 3: where it ships from
Number of despatch points and where they sit. This is the most expensive decision on the list because it ends in a lease, and because it fixes four separate costs at once: the outbound distance, the inbound distance, the stock you have to hold and the rate you will be quoted, since a site's return-load availability is what makes carriers cheap or dear to it. Where to put a warehouse so trucking costs stay low covers the arithmetic.
The reason it comes third and not first is that a site is chosen against a delivery map and a service promise. Choose it before decisions 1 and 2 and you are choosing against a guess.
Decision 4: the vehicle profile
Now the drop size distribution earns its keep. It tells you which vehicle sizes the work actually needs, and in what mix. Most operations discover two things here: that the truck everyone talks about is sized for the largest drops rather than the typical one, and that a small number of oversized drops is driving the whole conversation. Choosing the right truck size covers how volume, weight, loading access and temperature control interact.
Take the same operation as before, 18 drops on an average day, and sort its history by drop weight:
| Drop weight | Share of drops |
|---|---|
| Under 1,000 kg | 58% |
| 1,000 to 3,000 kg | 27% |
| 3,000 to 7,000 kg | 11% |
| Over 7,000 kg | 4% |
The conversation in the room will be about the 4%, because those are the deliveries people remember. The table says 85% of drops are under three tonnes.
Turn that into vehicles rather than opinions. If a smaller vehicle covers everything up to three tonnes, an average day is 18 × 0.85, or 15.3 small drops, and 18 × 0.15, or 2.7 larger ones. At six drops a vehicle-day that is 15.3 ÷ 6, which rounds to three small vehicles, and one larger vehicle with room to spare. The profile is three small and one large, not four of the same thing.
Two warnings on that. A mixed profile buys a lower cost per drop and gives up flexibility, because a small vehicle cannot cover for a large one when the large one is off the road. And the 4% deserves its own question rather than a bigger truck: drops that size are often better handled as a separate movement than carried as a constraint on every round for the rest of the year.
One Thai fact belongs here rather than later, because it turns a vehicle choice into a legal one.
A vehicle weight of 2,200 kilogrammes is the line that decides which law you are under. Under the Land Transport Act, การขนส่งส่วนบุคคล, private transport, means carrying goods for your own trade or business in a vehicle whose weight exceeds 2,200 kilogrammes. Below that, a private truck registered under the motor vehicle law is outside the Act altogether, but only if it is not being used to carry for hire. Both conditions have to hold. The Act's own definition of the for-hire categories turns on the words เพื่อสินจ้าง, for reward, so the moment money changes hands for the carriage itself, the size of the vehicle stops saving you.
Note the measure. This threshold is the weight of the vehicle, not the vehicle plus its load.
Decision 5: the legal frame you just chose
You did not pick this. Decision 4 picked it for you, which is why it sits here and not at the top.
If your vehicles are over 2,200 kilogrammes and you carry your own goods, you need an operating licence to do it. This is the part that surprises people. Section 23 of the Act prohibits fixed-route transport, non-fixed-route transport, transport by small vehicle and private transport without a licence from the registrar. Carrying your own goods in your own trucks is not exempt from licensing; it has its own licence. A breach of that section carries imprisonment of up to five years, or a fine of 20,000 to 100,000 baht, or both.
It is also completely normal. As at 28 February 2026 the Department of Land Transport's own figures show 312,558 private transport operating licences on issue for goods vehicles, against 38,071 non-fixed-route licences. Private licences outnumber for-hire ones by more than eight to one, and are about 89% of all goods-transport operating licences. The unit is the licence document rather than the company, and licences run five years, so the stock reflects renewals as well as new entrants. On the same series three years earlier there were 335,953 private goods licences, so the count has fallen about 7% while the for-hire count rose about 2%.
Four points that decide your paperwork, all of them read off decision 4:
The licence you need depends on who owns the goods, not on the truck. There are four operating licences: fixed route, non-fixed route, small vehicle and private transport. Carrying your own goods is the private transport licence. Carrying anyone else's for money is the non-fixed-route licence. The two are not symmetric: a fixed-route, non-fixed-route or small-vehicle licence may be used as a private transport licence, but a private one cannot be used the other way, and using a vehicle for the wrong type needs occasional written permission from the registrar. If there is any chance you will one day sell a backhaul to the factory next door, that is a different licence, and it is better decided now than after the trucks are marked.
The nationality test is written for the for-hire licences. Section 24 requires Thai nationality of the applicant, and for a company: registration under Thai law, a head office in the Kingdom, not less than half the directors of Thai nationality and not less than 51% of the capital held by Thai natural persons. It names the fixed-route, non-fixed-route and small-vehicle licences. It does not name private transport. For a foreign-owned manufacturer that plans to move only its own goods, that distinction is worth checking early with your own advisers, because it is the Land Transport Act's test and not the whole of Thai law on foreign ownership.
The licence names your trucks and your yard. When the registrar issues a private transport licence, Section 34 requires conditions to be written into it covering the number of vehicles used in the operation, their characteristics, type and size and the operator's mark shown on each one, the place where they are kept, repaired and maintained, and the number of crew. That is decision 3 and decision 4 written onto a government document. Apply before you have settled them and you are applying with numbers you will have to go back and change. Section 39 then bars you from using the licensed vehicle, or letting anyone else use it, outside the route or the locality authorised, as the case may be, without written permission from the registrar.
There is a second threshold, and it is measured differently. The driver's licence class turns on 3,500 kilogrammes of vehicle weight and load weight combined. Class one covers a goods vehicle at or under that figure; class two covers one above it. Class three is for towing and class four for dangerous goods, and the higher classes can be used as the lower ones but never the reverse.
| Threshold | What it measures | What it decides |
|---|---|---|
| 2,200 kg | Vehicle weight alone | Whether the Land Transport Act applies and an operating licence is needed |
| 3,500 kg | Vehicle weight plus load weight | Which driver licence class the person at the wheel must hold |
Two numbers, two different measures, and mixing them up is how an operation ends up correctly licensed with drivers who are not.
The last legal input to the plan is the driving day. Section 103 bis limits a driver to four continuous hours from the moment of starting, extendable by four more after an unbroken rest of not less than half an hour, within a twenty-four hour cycle. It caps hours at the wheel, not distance. Whatever route pattern comes out of the next decision has to fit inside it on the worst day, not the average one.
Decision 6: how much capacity you commit
Now, and not before, you can answer how much transport capacity to hold on a standing basis. The day shape from decision 1 is the input, and the arithmetic is short.
Take an operation whose history shows 18 drops on an average day and 31 on the busiest, in an area where one vehicle manages six drops a day.
- Sized to the peak: 31 ÷ 6 rounds up to six vehicles. On an average day those six carry 18 drops between them, which is 18 ÷ 36, or 50% of what they could do.
- Sized to the average plus 15%: 18 × 1.15 is 20.7, which rounds up to four vehicles. On an average day that is 18 ÷ 24, or 75%.
Four vehicles cover 24 drops. Every day in the history above 24 drops needs something brought in. Count those days in your own table rather than guessing: if it is nine days in sixty, you are buying help on 15% of days and running at 75% the rest of the time, instead of running at 50% all year to be ready for nine days.
The general principle behind it is old and holds up: resource the operation to somewhere between average demand and average plus 10 to 20 per cent, and cover the peaks with capacity you bring in rather than capacity you keep. It applies to the size of a committed carrier contract exactly as it applies to a vehicle count, because the thing being sized in both cases is standing capacity you pay for whether or not it moves.
Who supplies those vehicles is the next decision, and it is genuinely the next one. It cannot be answered honestly until you can hand somebody a drop profile, a service promise, a base volume and a peak volume, which is precisely what decisions 1 to 6 produce. Answer it first, as most people do, and you are asking a supplier to size something you have not measured.
Decision 7: the records, and the site detail
Two pieces of unglamorous work that belong here because their volume is only knowable once you know which sites you serve and how many vehicles you run.
Section 35 requires operators of a type announced by the Director-General to keep a vehicle log book, crew records, records of the inspection of vehicle and driver readiness with those inspections recorded, and the consignment note for the goods being carried, together with transport reports and accident reports. Decide who holds these and where before the first vehicle moves, because reconstructing three months of them afterwards is a job nobody ever finishes.
The site detail is the other one: for each delivery point, the access, the vehicle size it can physically take, the opening hours, and the restrictions on the roads leading to it. This is per-site work, so its cost scales with the number of points and not with anything else, and it is the single most common reason a plan that looked right on a spreadsheet fails in its first week.
Decision 8: the baseline, taken before you change anything
Take the measurements while the operation is still doing whatever it did before. Cost per drop, cost per delivered tonne or case, the day shape, and on-time performance against the promise from decision 2.
If you take them after the changes, you have no way of telling anybody what the changes were worth, including yourself. This is the cheapest step on the list and it is the one with a deadline, because the chance to measure the old operation disappears the day the new one starts.
The order you decide is not the order you start
One thing gets missed by everyone who reads a list like this and treats it as a schedule. The decisions are ordered by dependency. The work is ordered by how long it takes, and the two are not the same.
Three of these steps consume calendar time you cannot compress:
- The site search. Months, and it is on the critical path from day one.
- The despatch history. Eight to twelve weeks of it, though if it already exists in a system you are extracting rather than waiting.
- Licensing and registration. Slow rather than hard, and it cannot start until decisions 3 and 4 are settled, because the licence has to name the vehicles and the place they are kept.
So on day one you start decision 1 and the site search together. They do not depend on each other and both are long. Everything else waits for a real input rather than being started early to look busy.
The common failure is exactly that: filing the licence application early to get it out of the way, then amending it twice when the vehicle count and the yard change. Amendments are routine, but each one is a trip back to the registrar, and a licence that says four vehicles is not a licence for six.
The order, and what each step costs to undo
| # | Decision | Needs from the step before | Cost to reverse in month six |
|---|---|---|---|
| 1 | What you already ship | Nothing. Start here | None. Redo it any time |
| 2 | What you promise | The drop and day profile | High. It sits in customer contracts |
| 3 | Where it ships from | The delivery map and the promise | Highest. It sits in a lease |
| 4 | The vehicle profile | The drop size distribution | Moderate. Vehicles can be changed at renewal |
| 5 | The legal frame | The vehicle profile and the site | Moderate, and slow. Licence conditions are amended, not ignored |
| 6 | How much you commit | The day shape and the vehicle profile | Moderate. Contract length decides it |
| 7 | Records and site detail | Which sites, how many vehicles | Low, but the gap is never filled retrospectively |
| 8 | The baseline | Everything still being as it was | Zero now, impossible later |
If you have already started in the wrong order
Most people reading this have. The operation is running, the site is leased, the promises are made, and the trucks are booked.
The recovery is not to redo the sequence. It is to do decision 1 properly, which nobody can stop you doing, and then check each downstream decision against what the table actually says. Almost always the same two things fall out: the vehicle profile is sized for the largest drop rather than the typical one, and the committed capacity is sized for the peak day rather than for the average plus a margin. Both are fixable at the next renewal, and both are invisible until the table exists.
Start there. Eight weeks of despatch history, five columns.